Research On Intelligent Semantic Mapping Based On Distributed Description For Indoor Robots

With the unceasing expansion of robot application, highly intelligent robot development has become an inevitable trend. Understanding the semantic features of the environment is a premise of achieving human-robot interaction, reducing the perception gap and performing complicated tasks. However, conventional map lacks understanding of the environmental semantic information, and thus seriously hampers its theoretical development and practical application. Considering this problem, and aiming at the indoor work environment, this paper studies an intelligent semantic mapping method based on distributed description for semantic characterization of the scene. With the semantic information embedded in the artificial landmarks, the robot acquires the semantic information by scanning the landmarks, and then establishes a semantic map for providing guidance on the follow-up work. The main contents of this paper include:1. For indoor environment, this paper studies a solution to SLAM problem for a mobile robot with binocular vision sensors. The two-wheeled mobile robot shoots the environmental images by the binocular vision sensor, and computes the depth information of landmark using parallax method. The robot coordinates are optimally estimated by EKF algorithm, and the environmental landmarks are positioned and propagated at the same time.2. Considering influence of some problems, such as illumination change and horizon limitation, on the visual localization results, RFID system is arranged in the indoor environment. The coordinates of the artificial landmarks and the observation points of the mobile robot with electronic label are both calculated based on BVIRE algorithm. RFID localization results can modify the errors of visual localization to significantly improve the localization accuracy of the robot.3. For the problem of semantic cognizance and understanding for mobile robot in indoor environment, we design a kind of QR code artificial landmarks, which are convenient for semantic information to be stored and read. And a metric-topological-semantic distributed intelligent mapping method based on QR code technology and RFID technology is proposed. In metric layer, the robot optimizes its position and the position of artificial landmarks using visual information and RFID information. In topological layer, the coordinates of robot observation points are segmented on the basis of the images taken by binocular camera to construct the topological map. In semantic layer, with scanning the QR code with semantic information, the robot divides the indoor environment according to function. The semantic information endows and updates the semantic attributes of the nodes of topological map, and at last a semantic map is established.This paper researches a method of intelligent semantic mapping based on distributed description for indoor robots. The method provides the environmental cognizance for robots not only on the physical level, but also on the semantic level, and lays the foundation for the robot to complete the complex indoor tasks.